U.S. patent number 10,060,726 [Application Number 15/654,297] was granted by the patent office on 2018-08-28 for parallel image measurement method oriented to the insulating layer thickness of a radial symmetrical cable section.
This patent grant is currently assigned to SHANGHAI UNIVERSITY OF ENGINEERING SCIENCE. The grantee listed for this patent is SHANGHAI UNIVERSITY OF ENGINEERING SCIENCE. Invention is credited to Xiang Liu, Yunyu Shi, Yongxiang Xia.
United States Patent |
10,060,726 |
Liu , et al. |
August 28, 2018 |
Parallel image measurement method oriented to the insulating layer
thickness of a radial symmetrical cable section
Abstract
The present invention relates to a parallel image measurement
method oriented to the insulating layer thickness of a radial
symmetrical cable section. The method conducts the non-contact
high-accuracy measurement based on the machine vision and the image
analysis, adopts a GPU multi-core parallel platform for the
high-speed measurement, extracts the useful information from the
section image of the radial symmetrical cable, and then measures
the insulating layer thickness. Compared with the prior art, the
present patent can lower the time consumed for the accurate
measurement, fill in the blank of the high-accuracy parallel image
measurement of the insulating layer thickness of the radial
symmetrical cable section in the domestic cable industry, break
down the monopoly and technology blockade by related foreign
manufacturers and improve the technology level of on-line testing
of product quality in China, expedite the production automation
progress of domestic manufacturer.
Inventors: |
Liu; Xiang (Shanghai,
CN), Shi; Yunyu (Shanghai, CN), Xia;
Yongxiang (Shanghai, CN) |
Applicant: |
Name |
City |
State |
Country |
Type |
SHANGHAI UNIVERSITY OF ENGINEERING SCIENCE |
Shanghai |
N/A |
CN |
|
|
Assignee: |
SHANGHAI UNIVERSITY OF ENGINEERING
SCIENCE (Shanghai, CN)
|
Family
ID: |
54663473 |
Appl.
No.: |
15/654,297 |
Filed: |
July 19, 2017 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20180017375 A1 |
Jan 18, 2018 |
|
Related U.S. Patent Documents
|
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
Issue Date |
|
|
PCT/CN2016/072495 |
Jan 28, 2016 |
|
|
|
|
Foreign Application Priority Data
|
|
|
|
|
Apr 24, 2015 [CN] |
|
|
2015 1 0197679 |
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06T
7/12 (20170101); H04N 5/232 (20130101); H04N
5/142 (20130101); H04N 5/372 (20130101); G06T
7/149 (20170101); G06T 7/0004 (20130101); G06T
7/62 (20170101); G01B 11/06 (20130101); G06T
2215/16 (20130101) |
Current International
Class: |
G01B
11/06 (20060101); H04N 5/372 (20110101); H04N
5/14 (20060101) |
References Cited
[Referenced By]
U.S. Patent Documents
Foreign Patent Documents
|
|
|
|
|
|
|
101206109 |
|
Jun 2008 |
|
CN |
|
102032875 |
|
Apr 2011 |
|
CN |
|
102226687 |
|
Oct 2011 |
|
CN |
|
102359761 |
|
Feb 2012 |
|
CN |
|
103336959 |
|
Oct 2013 |
|
CN |
|
204142186 |
|
Feb 2015 |
|
CN |
|
105115428 |
|
Dec 2015 |
|
CN |
|
2007115461 |
|
May 2007 |
|
JP |
|
2009032516 |
|
Feb 2009 |
|
JP |
|
Other References
International Search Report (Chinese and English) and Written
Opinion of International Application No. PCT/CN2016/072495, dated
Apr. 27, 2016, 13 pages. cited by applicant.
|
Primary Examiner: Nguyen; Sang
Attorney, Agent or Firm: Hamre, Schumann, Mueller &
Larson, P.C.
Claims
What is claimed is:
1. A parallel image measurement method oriented to an insulating
layer thickness of a radial symmetrical cable section, the method
comprising: an industrial CCD camera calibrating an image of the
radial symmetrical cable section; conducting a non-contact
high-accuracy measurement based on a machine vision and image
analysis; a multi-thread multi-core processor performing a
high-speed measurement; obtaining information from the image of the
radial symmetrical cable section; and measuring the insulating
layer thickness; wherein measuring the insulating layer thickness
includes: obtaining an inner and an outer contour of the radial
symmetrical cable section from the image and calculating a mass
center of the cable section; pinpointing pixels of the inner
contour to sub-pixels, connecting the mass center and the pixels of
the inner contour and prolonging to the outer contour; piece-wise
curve fitting the outer contour, obtaining a junction of the outer
contour and an elongation line; obtaining a distance between the
junction and the pixels of the inner contour, the distance being
the insulating layer thickness corresponding to the pixels of the
inner contour; and obtaining a maximum thickness, a minimum
thickness, and an average thickness of the insulating layer of the
radial symmetrical cable section.
2. The parallel image measurement method according to claim 1,
wherein pinpointing the pixels of the inner contour and piece-wise
curve fitting the outer contour are implemented by the multi-thread
multi-core processor, the multi-thread multi-core processor
implementing a B-spline curve fitting via parallel calculation.
3. The parallel image measurement method according to claim 2,
wherein the multi-thread multi-core processor implementing the
B-spline curve fitting includes: {circle around (1)} starting the
multi-thread multi-core processor, allocating space in a display
memory, and copying data to the display memory; {circle around (2)}
defining a number of blocks and threads, spawning the threads,
calling a kernel function, adopting the B-spline curve fitting to
realize the pinpointing of the inner contour; {circle around (3)}
defining the number of the blocks and the threads, spawning the
threads, calling the kernel function, and using the mass center,
points on the inner contour and a fitting function to obtain
corresponding points on the outer contour; {circle around (4)} the
display memory and the multi-thread multi-core processor
transferring a calculated result to a CPU, and releasing resources
on the display memory and the multi-thread multi-core
processor.
4. The parallel image measurement method according to claim 1,
wherein a value is identified from all candidate values of
thickness as a final measured value to obtain the maximum and the
minimum thickness.
5. The parallel image measurement method according to claim 4,
wherein obtaining the minimum thickness comprises: (1) sequencing
the obtained thickness of the insulating layer corresponding to
each of the pixels of the inner contour in an ascending order; (2)
taking N minimal values and corresponding 2D co-ordinates, marking
the N minimal values and the 2D co-ordinates as set Tn; (3) for an
ith minimum value of the N minimal values, defining a weight Wi=0;
if a point q adjacent to the ith minimum value in the image is in
Tn, and a sequencing interval of the ordered thickness does not
exceed 10, then Wi++, and inactivating the adjacent point in TN;
(4) performing steps (1)-(3) for all of the N minimum values in the
ascending order, and if the Wi is greater than a certain threshold,
a current Ti is the minimum thickness.
6. A parallel image measurement system for an insulating layer
thickness of a radial symmetrical cable section, the system
comprising: an industrial CCD camera, the camera is configured to
calibrate an image of the radial symmetrical cable section; a CPU;
and a multi-thread multi-core processor, wherein the CPU and the
multi-thread multi-core processor are configured to conduct a
non-contact high-accuracy measurement based on a machine vision and
image analysis; perform a high-speed measurement; obtain
information from the image of the radial symmetrical cable section;
and measure the insulating layer thickness, wherein when the CPU
and the multi-thread multi-core processor measure the insulating
layer thickness, the CPU and the multi-thread multi-core processor
are further configured to: obtain an inner and an outer contour of
the radial symmetrical cable section from the image and calculate a
mass center of the cable section; pinpoint pixels of the inner
contour to sub-pixels, connect the mass center and the pixels of
the inner contour and prolong to the outer contour; piece-wise
curve fit the outer contour, obtain a junction of the outer contour
and an elongation line; obtain a distance between the junction and
the pixels of the inner contour, the distance being the insulating
layer thickness corresponding to the pixels of the inner contour;
and obtain a maximum thickness, a minimum thickness, and an average
thickness of the insulating layer of the radial symmetrical cable
section.
7. The parallel image measurement system of claim 6, wherein the
multi-thread multi-core processor is configured to pinpoint the
pixels of the inner contour and piece-wise curve fit the outer
contour, and the multi-thread multi-core processor is configured to
implement a B-spline curve fitting via parallel calculation.
8. The parallel image measurement system of claim 7, further
comprising: a display memory, wherein when the multi-thread
multi-core processor implements the B-spline curve fitting, the
multi-thread multi-core processor is started, and the multi-thread
multi-core processor and the display memory are configured to:
allocate space in the display memory and copy data to the display
memory; define a number of blocks and threads, spawn the threads,
call a kernel function, adopt the B-spline curve fitting to
pinpoint the inner contour; define the number of the blocks and the
threads, spawn the threads, call the kernel function, and use the
mass center, points on the inner contour, and a fitting function to
obtain corresponding points on the outer contour; and transfer a
calculated result to the CPU, and release resources on the display
memory and the multi-thread multi-core processor.
9. The parallel image measurement system of claim 6, wherein the
CPU and the multi-thread multi-core processor are configured to
identify a value from all candidate values of thickness as a final
measured value to obtain the maximum and the minimum thickness.
10. The parallel image measurement system of claim 9, wherein when
the CPU and the multi-thread multi-core processor obtain the
minimum thickness, the CPU and the multi-thread multi-core
processor are configured to: (1) sequence the obtained thickness of
the insulating layer corresponding to each of the pixels of the
inner contour in an ascending order; (2) take N minimal values and
corresponding 2D co-ordinates, mark the N minimal values and the 2D
co-ordinates as set Tn; (3) for an ith minimum value of the N
minimal values, define a weight Wi=0; if a point q adjacent to the
ith minimum value in the image is in Tn, and a sequencing interval
of the ordered thickness does not exceed 10, then Wi++, and
inactivate the adjacent point in TN; (4) perform (1)-(3) for all of
the N minimum values in the ascending order, and if the Wi is
greater than a certain threshold, a current Ti is the minimum
thickness.
Description
BACKGROUND OF THE INVENTION
Technical Field
The present invention relates to the field of machine vision
industrial testing, especially relates to a parallel image
measurement method oriented to the insulating layer thickness of a
radial symmetrical cable section.
Description of Related Art
The industry of electrical wire and cable is the second largest
industry next only to the automobile industry in mechanical
manufacturing. It is estimated that as China will enter the later
stage of industrialization in the next several years, the rate of
development of the electrical wire and cable industry in China will
be higher than that of the national economy, estimated to be over
10%, while the average annual increase for the electrical conductor
and cable can be up to 15%. As the core of many power
infrastructures, the cable determines the security and reliability
of the whole power grid. Once the product quality deviation appears
in the production line of a cable and fails to be timely adjusted,
substantive defective products will be produced, leading to a lot
of waste. Once the defective products enter the market, they will
cause astronomically more power loss. Therefore the structure
measurement of a cable is an essential link in product quality
monitoring system, also serves as an important factor to guarantee
the enterprises outperformance in the cut-throat market
competition. The structural measurement of electric wire and cable
has experienced the first-generation caliper, the second-generation
micrometer reading microscope measurement and the third-generation
digital projector. The traditional measurement methods can only
measure limited points and local parts with a lot of missing
points, thus making it hard to guarantee accuracy. Besides, the
time consuming and laborious manual operations just produce
unstable measurement effects that are significantly related to
operators.
In recent years, the foreign advanced countries have introduced the
high definition (HD) industrial camera, which is based on the
industrial vision theory and technology to explore high-accuracy
image measurement, with borderline products being put in commercial
application gradually. However, the equipment is expensive and
cannot be adjusted to the actual conditions of the domestic
manufacturers. Market research displays an estimated total demand
of more than 1,000 units, and similar foreign equipment is priced
at 500,000-800,000 Yuan per unit. From the perspective of social
value, an automatic optic inspection (AOI) system can operate
continuously for long with a strong adaptability to the tough work
environment, thus avoiding the negative influence from continuous
high-strength work of the workers, thereby providing a guarantee
for humanistic management of the enterprises. From the perspective
of national interests, currently the Chinese AOI technology based
on the computer vision has just started, with the main application
fields of low speed and low precision such as character
recognition, inspection of printing quality and product selecting,
etc., while there is still dependence on the foreign products in
high-precision and high-speed inspection, with costly purchase and
maintenance expenses. Some home-made inspection systems based on
the computer vision also adopt the core foreign technology. These
products' key technologies such as system structure and core
algorithm are still in the hands of foreign manufacturers, who
often provide us with outdated technologies about to be washed
out.
The commercially available cables have a lot of radial symmetrical
sections. To eliminate the defects in the traditional measurement
methods, the major technology problems encountered in the real-time
calculation of the minimum thickness, the average thickness and the
maximum thickness of insulating layer comprise:
(1) how to improve the measurement accuracy of the section
image
(2) how to improve the computing speed of the complicated image
BRIEF SUMMARY OF THE INVENTION
The purpose of the present invention is to provide a parallel image
measurement method oriented to the insulating layer thickness of a
radial symmetrical cable section to overcome the above defects of
prior art. This method can help the cable quality inspectors to
non-contact measure the insulating layer thickness of radial
symmetrical cable section in real time during quality inspection,
and can analyze and infer the problems existing in the production
or processing of the cable by dint of measurement data.
The purpose of the present invention can be achieved by the
following technical solution:
A parallel image measurement method oriented to the insulating
layer thickness of a radial symmetrical cable section,
characterized in that, the method conducts the non-contact
high-accuracy measurement based on the machine vision and the image
analysis, adopts a GPU multi-core parallel platform for the
high-speed measurement, extracts the useful information from the
section image of the radial symmetrical cable to measure insulating
layer thickness.
Specifically, the method comprises the following steps:
1) reading an image shot, calibrated by an industrial CCD
camera;
2) extracting an inner and an outer contour of the radial
symmetrical cable section from the image, and calculating a mass
center of the cable section;
3) subjecting the pixels in the inner contour to the sub-pixel
pinpointing, connecting the mass center and the pixels of the inner
contour and prolonging to the outer contour;
4) subjecting the outer contour to the piece-wise curve fitting,
and solving a junction of the outer contour and an elongation
line;
5) calculating the distance between the junction and the pixels of
the inner contour, which will be the insulating layer thickness
corresponding to the current pixels of the inner contour;
6) adopting a statistical method to obtain the maximum thickness,
the minimum thickness and the average thickness of the insulating
layer of this radial symmetrical cable section
Realizing the B-spline curve fitting method based on a GPU
multi-core parallel calculation platform, to realize pinpointing of
the inner contour pixels and piecewise fitting of the outer
contour.
Specifically, realizing B-spline curve fitting method based on the
GPU multi-core parallel calculation platform is done by:
{circle around (1)} starting a GPU, allocating space in a display
memory and copying data to the display memory;
{circle around (2)} defining the number of the blocks and the
threads, spawning the threads, calling a kernel function, adopting
the B-spline curve fitting to realize sub-pixels pinpointing of the
inner contour points;
{circle around (3)} defining the number of the blocks and the
threads, spawning the threads, calling the kernel function,
calculating with the mass center, the points on the inner contour
and the fitting function, to obtain the corresponding points on the
outer contour;
{circle around (4)} the display memory and the GPU transfer the
calculated results to a CPU, the resources on the display memory
and the GPU are released.
Adopting the statistical method to obtain the maximum thickness,
the minimum thickness and the average thickness of the insulating
layer of this radial symmetrical cable section, looking for an
appropriate value from all candidate thickness values as the final
measured value to solve the maximum and the minimum values of the
thickness, are not using a simple sorting algorithm. If an inner
wall point corresponding to the thickness extrema is an active
pixel, then there will exist heaps of similar active pixels around
this pixel, making the neighboring thickness value an approximation
of the thickness extrema.
The specific steps in solving the minimum thickness are as
follows:
(1) sequencing the calculated thickness value of the insulating
layer corresponding to each pixel of inner contour in an ascending
order.
(2) taking N minimum values and the corresponding 2D co-ordinates,
marking them as set Tn;
(3) for the ith minimal value, defining the weight Wi=0; if the
point q adjacent to it in the image is in Tn, and the sequencing
interval of the thickness does not exceed 10, then Wi++, and
letting the adjacent points inactive in Tn;
(4) making the same operation for N minimum values in the ascending
order. If Wi is greater than a certain threshold, the current Ti is
the minimum value.
The present invention contains the following benefits over the
prior art
Firstly, with respect to characteristics of radial symmetrical
cable section, this method uses the section image, offers a
parallel measurement technology solution suitable for non-contact
measurement and cuts the time consumed for the accurate measurement
via the GPU multi-core parallel calculation platform.
Secondly, it pinpoints the pixels of the inner contour of cable
through an analysis on the section image of the cable, adopts the
B-spline curve fitting to piecewise fit the outer contour, thereby
laying a foundation for the high-accuracy measurement.
Thirdly, it looks for an appropriate value from all the candidate
thickness values measured as the last measurement value, thereby
eliminating the influence of inactive pixels on the measurement
accuracy.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS
FIG. 1 is a technical line map of the present invention;
FIG. 2 is an organizational structure schematic diagram of the
threads;
FIG. 3 is a schematic diagram of a cable section.
DETAILED DESCRIPTION OF THE INVENTION
The present invention will now be described in detail in connection
with an embodiment of the present invention with reference to the
accompanying drawings.
Embodiment
The difference from the existing measurement methods resides in
that the present invention conducts the non-contact high-accuracy
measurement based on the machine vision and the image analysis,
adopts a GPU multi-core parallel platform for the high-speed
measurement. It mainly considers how to extract the useful
information from the section image of the radial symmetrical cable,
and then measures the insulating layer thickness systematically, as
shown in FIG. 1. The specific steps of the present invention are as
follows:
{circle around (1)} reading an image shot, calibrated by an
industrial CCD camera;
{circle around (2)} extracting an inner and an outer contour of a
radial symmetrical cable section from the image, and calculating a
mass center of the cable section;
{circle around (3)} subjecting the pixels of the inner contour to
the sub-pixel pinpointing, connecting the mass center and the
pixels of the inner contour and prolonging to the outer
contour;
{circle around (4)} subjecting the outer contour to the piece-wise
curve fitting and solving a junction of the outer contour and an
elongation line;
{circle around (5)} calculating the distance between the junction
and the pixels of the inner contour, which will be the insulating
layer thickness corresponding to the current pixels of the inner
contour;
{circle around (6)} adopting a statistical method to obtain the
maximum thickness, the minimum thickness and the average thickness
of the insulating layer of this radial symmetrical cable
section
The technology solution of the present invention is used by the
quality inspectors of cable to obtain the measurement data of the
insulating layer thickness via parallel calculation of the section
image, specifically including the following:
1. Pinpointing of the pixels of the inner contour and piecewise
fitting the outer contour based on a GPU platform
(1) Rationale
In consideration of the calculation capacity and performance of the
hardware, a GPU parallel calculation model can be adopted to
overcome the problems of too much time consumed and slow speed in
the procedure of testing cable thickness, to enable a dramatic
reduction of the time consumed in the testing procedure. The GPU
(Graphics Processing Unit) is a highly parallel and multi-threading
multi-core processor with a powerful computing capacity and a high
band width. GPU parallel calculations can improve the performance
of image processing dozens of times.
The CUDA is a soft hardware system with a GPU as the data parallel
calculation equipment, developed by C language and easy for
learning and use. A CPU serves as a host to do strongly logical
tasks and serial computing. The GPU serves as a device to do highly
threaded parallel processing tasks. Adopting a CPU+GPU isomerical
parallel processing can significantly lower the burden to the CPU,
decrease the CPU system overhead, raise the whole throughput of the
system, improve the computing capacity of the system and economize
on the cost and energy resources. A GPU parallel algorithm is
adopted in the procedure of testing cable thickness to overcome the
problems of too much time consumed and slow speed of calculation.
The organizational structure of the threads is as shown in FIG.
2.
(2) Basic Steps
The B-spline curve fitting method can only target one certain pixel
on the contour in each calculation, and the one calculated is a
sub-pixel edge position of an individual pixel. The sub-pixel
positioning of the edge entails calculating whole pixels on the
contour one by one, with higher positioning accuracy yet relatively
slow speed. Besides, as the image adopts a resolution of 4092*4092,
the quantity of pixels on the inner and outer contours becomes very
large and the calculation consumes a lot of time. Adopting the GPU
parallel calculation model can considerably shorten the time
consumed in this procedure. The main steps are as follows:
{circle around (1)} starting the GPU, allocating space in a display
memory and copying the data to the display memory;
{circle around (2)} defining the number of the blocks and the
threads, spawning the threads, calling a kernel function, adopting
a B-spline curve fitting to realize sub-pixels pinpointing of
contour points;
{circle around (3)} defining the number of the blocks and the
threads, spawning the threads, calling the kernel function,
calculating the mass center, the points on the inner contour and
the fitting function, to obtain the corresponding points on the
outer contour;
{circle around (4)} the display memory and GPU transfer the
calculated results to the CPU and the resources on the display
memory and GPU are released.
Mean and extrema calculation of insulating layer thickness based on
the statistical method
(1) Rationale
Every cable section has an objective true value, and the most ideal
measurement is to get this true value. However, the cable section
is measured by humans using a CCD camera under certain
illumination, which is limited by the sensitivity and the
resolution capacity of the camera as well as the environmental
instability, etc., hence the true value to be measured is
immeasurable. Therefore, due to the natural limitation of accuracy
and precision of the CCD camera, there are still residual inactive
pixels even if the image is de-noised. The accuracy of the measured
value will be influenced if the inactive pixel coincides with the
inner wall point corresponding to the thickness extrema.
Thus an appropriate value should be selected from all candidate
thickness values as the last measured value to solve the maximum
value and the minimum value of the thickness, instead of using a
simple sorting algorithm. If the inner wall point corresponding to
the thickness extrema is an active pixel, then there will exist
heaps of similar active pixels around this pixel, making the
neighboring thickness value only an approximation of the thickness
extrema.
(2) Basic Steps
Calculate the maximum thickness, the minimum thickness and the
average thickness of the insulating layer of the radial symmetrical
cable section, i.e. select the appropriate value from all candidate
thickness values as the last measured value. The schematic diagram
of the radial symmetrical cable is as shown in FIG. 3. The specific
steps of solving the minimum value of the insulating layer
thickness include the following:
{circle around (1)} sequencing the calculated thickness values of
the insulating layer corresponding to each pixel of inner contour
in an ascending order.
{circle around (2)} taking N minimum values and corresponding 2D
co-ordinates, marking them as set Tn;
{circle around (3)} for the ith minimal value, defining the weight
Wi=0; if the point q adjacent to it (point distance smaller than 3)
in the image is in Tn, and the sequencing interval of the thickness
does not exceed 10, then Wi++, and letting the adjacent points
inactive in TN;
{circle around (4)} making the same operation for N minimum values
in the ascending order. If Wi is greater than a certain threshold,
the current Ti is the minimum value.
The present invention enables real-time calculation of the minimum
size, the maximum size and the average size of the insulating layer
thickness after obtaining the image of the cable cross section
scanned by an HD industrial camera in a full-coverage way. The
present patent will fill in the blanks of the high-accuracy
parallel image measurement of the insulating layer thickness of the
radial symmetrical cable section in the domestic cable industry,
break down the monopoly and technology blockade by the concerned
foreign manufacturers and improve the technology level of on-line
measurement of product quality in China. Furthermore, it can
expedite the production automation progress of domestic
manufacturers, economize a great deal on labor, financial resources
and material resources. The potential application is wide and
expandability is satisfactory. In addition, the technology can be
further developed to be applied to high-accuracy image measurement
of enamel wire structures.
* * * * *